Murine CRC cell lines MC-38 (MMR-d, MSI) were kindly provided by Dr. Pnina Brodt, McGill University, who obtained them from Dr. Shoshana Yakar, New York University. CT-26 (MMR-p, MSS) were generously provided by Dr. Nicole Beauchemin of McGill University, sourced directly from Dr. Michael Brattain of the University of Nebraska Medical Centre. Dr. Brattain originally established the cell line in 1980 and granted permission to Dr. Beauchemin to share this cell line with other investigators (41). The human CRC cell lines HT-29 (MMR-p, MSS) (ATCC HTB-38) and HCT116 parental cells (KRAS^G13D/KRAS^WT, ATCC CCL-247, Horizon Cat. HD PAR-007) and HCT116 WT (KRAS ^WT/KRAS ^KO, cat. HD-104-008), and HCT116 KRAS^G13D (KRAS^G13D/KRAS ^KO, cat. HD 104-011) (MMR-d) were obtained from Horizon Discovery, Ltd. Growth conditions can be found in the Supplementary Material .

Cytotoxicity was measured using the sulforhodamine B assay as previously described (42).

Cells were detached with 0.05% trypsin-EDTA and seeded at 3 × 10³ cells per well in 96-well plates. 24h later, the cells were treated with serial dilutions of copper (II) diethyldithiocarbamate (TCI Chemicals) reconstituted in DMSO (Sigma). DMSO concentration in cell suspension was normalized across all wells and did not exceed 0.1%. Cells were incubated at 37°C, 5% CO₂ for 72h. Cultures were fixed with 50% (w/v) trichloroacetic acid (Sigma) and stained for 30 min with 0.4% (w/v) sulforhodamine B (SRB; Alfa Aesar) dissolved in 1% (w/v) acetic acid (Sigma). Unbound dye was removed through four washes with 1% acetic acid solution, and protein-bound dye was extracted using 10 mM unbuffered Tris base (FisherScientific). OD492nm was measured. IC50 values were determined by calculating the percentage of cells killed in each test well as compared to the DMSO control well using the formula: % cells killed = 100 – (mean OD sample)/(mean OD control) x 100. IC50 values were determined by plotting a dose–response curve between the compound concentration and percent growth inhibition.

Clonogenic and migration-invasion assays’ methods can be found in the Supplementary Material .

Blots were incubated with i) primary antibodies against β-actin (MAB1501R, Millipore, 1:5000) paired with secondary horseradish peroxidase (HRP) antibody (#405306, BioLegend, 1:10000), ii) with primary antibodies anti-PARP (#9542S, 1:1000), and iii) anti-XIAP (#2045S, 1:1000) antibodies (Cell Signaling Technologies) paired with secondary HRP antibody (#31458, Invitrogen, 1:10000). For a detailed description, see Supplementary Material .

For in-vivo experiments, mouse albumin-based CuET nanoparticles (1 mg/mL) were synthesized as previously described (21). CuET powder is hardly soluble in any aqueous solution. Thus, for in-vivo experiments, mouse-albumin-based CuET nanoparticles (1mg/mL) were synthesized. Briefly, diethyldithiocarbamate (Sigma) (solved in water) is mixed with copper(ii) chloride (Sigma) (also in water) in a ratio 2:1. The reaction between these two compounds is carried in a 5% (w/v) mouse serum albumin (Innovative Research) solution. The resulting solution is stable at 4°C for at least 1 week. The dose administered to mice corresponded to the final concentration of 1 mg/kg of diethyldithiocarbamate-copper complex; the total injection volume was 0.5mL in isotonic salt.

8-week-old C57BL/6 and BALB/c mice received subcutaneous injections on the back, between the shoulder blades of 3.2 x 10⁵ MC-38 cells or 3 × 10⁵ CT-26 cells in 100μL PBS, respectively. Tumors became visible in the MC-38 and CT-26 grafts, ten- and seven-days following implantation on the backs of the mice, between the shoulder blades. Mice were randomized based on body weight into the following treatment groups: non-treated, vehicle (mouse albumin), and CuET (1 mg/kg). Treatment was administered through intraperitoneal injection (i.p.) as per the treatment schedules shown in respective figures. Tumor growth was monitored every second day through width (w) and length (L) measurements using an electronic calliper. Tumor volume (V) was calculated using the following formula: (V= W² × L). Body weights were monitored every second day. Endpoint tumor volume was defined as 2 cm³ according to the FACC guidelines. Survival of the tumor-bearing mice was evaluated using Kaplan-Meier analysis. For the metastatic model, 8-week old C57BL/6 mice of mixed sexes were anaesthetized with isoflurane, 20 mg/kg carprofen and 0.1 mg/kg regular buprenorphine (s.c.) 30 min prior to surgery. Mice were placed in the right lateral recumbent position. The spleen was exposed through a small flank incision immediately below the rib cage (0.5 cm). 5x10⁵ MC-38 tumor cells were inoculated with a 28Gx1/2 U-100 0.5cc insulin syringe into the spleen and after 1 minute, the splenic arteries and venous supply were cauterized, and the spleen was removed. The mice were sutured using 5-0 Vicryl sutures. Hepatic cell and splenocyte cytotoxicity were evaluated on day 6 post-implantation of tumor cells in a subset of mice, whereas metastatic burden was assessed on day 14 in another subset. For metastatic burden analysis, livers were measured, and the total area was calculated as follows: (A= L*W). Metastatic lesions were counted, and the area occupied by the lesions was calculated according to the radius using: (A=πr²). All experimental procedures were performed in accordance with the Facility Animal Care Committee of McGill University Health Center, Montreal, QC, Canada.

24h after the last injection, the internal organs and tumors from the CuET-treated and control mice were removed, fixed with 10% buffered formalin for 48h, embedded in paraffin, and sectioned at 4µm. Hematoxylin and eosin (H&E) staining was performed. Immunostaining was performed using an automated immunostainer (Discovery XT; Ventana Medical Systems). Tumor tissue sections were stained with anti-CD3 (#05493315001, Roche), anti-CD19 (AB245235, Abcam), anti-F4/80 (#70076, Cell Signaling), anti-neutrophil elastase (PA5-79198, ThermoFisher), anti-MICA (100507436-MSM2-P0, ThermoFisher), and anti-ULBP1 (17715-1-AP, ThermoFisher) to stain T cells, B cells, macrophages, neutrophils, and NK cell ligands on cancer cells MICA and ULBP1. Scanning was performed using a Leica Aperio AT Turbo digital pathology scanner at 40X magnification and 25 microns/pixel. The infiltrating cells were quantified in QuPath-0.3.2, using ten fields per slide, data expressed as % positive cells per field, n=3 per group.

For sorting, mouse hepatocytes from vehicle-treated mice and mice treated with 1 mg/kg CuET, were isolated in a 5.75 mL HBSS, 3.75 mL Percoll, 0.50 mL Heparin 1% in H₂O gradient following the previously published protocol and stained fresh (43). Cells were stained with Fixable Viability Dye eFluor 506 (Invitrogen, 65086614, 1:1000). Extracellular staining with anti-CD3 APC-eF780 (Invitrogen, 47-0032-82, 1:200), anti-CD8b APC-Cy7 (YTS156.7.7, Biolegend,126620, 1:200), and anti-NK1.1 FITC (PK136, Invitrogen eBiosciences, 11-5941-85, 1:150) was performed. For NK and NKT cell enrichment, NK cells were defined as NK1.1+ and CD3 −, whereas NKT cells were defined as NK1.1+ and CD3+. Sorting was performed using a BD FACSAria III instrument.

Human PBMC isolation and NK and T cell enrichment are described in detail in Supplementary Material . PBMC were incubated with autologous adherent cells and treated with or without 1nM CuET, to generate cytotoxic T cells, as previously described (44). Target tumor cells were maintained at 5 × 10⁴ cells. Four effector-to-target (E: T) ratios (20:1, 10:1, 5:1, and 2:1) were used for each assay. The Invitrogen™ CyQUANT™ LDH Cytotoxicity Assay Kit (ThermoFisher, C20301) was used to measure T and NK cytotoxic responses according to the manufacturer’s instructions. Lytic units were calculated using all four E:T ratios, as previously described (45).

Mouse organ harvesting and immunostaining was done as previously described (46). NK and T cells from human PBMC were first purified and enriched by negative selection as described in the Supplementary Material . After lymphocyte isolation, single-cell suspensions were stained with the following fluorescence-conjugated antibodies: NK and T cells untreated and treated with 1nM CuET were labeled with anti-human NKG2D (clone 1D11, ThermoFisher). Tumor cells were labelled with the NKG2D ligands: MICA (clone 159227, R&D Systems), ULBP1 (clone 170818, R&D Systems), and ULBP2/5/6 (clone 165903, R&D Systems). Antibody blocking experiments were performed using the same antibodies and the Thermo Scientific™ Pierce™ F(ab’)2 Preparation Kit (cat.44988). For perforin assessment from mouse splenocytes, the cells were fixed and permeabilized using the BDCytofix/Cytoperm™ Fixation/Permeabilization Kit (Cat.No.554714). Intracellular perforin staining with anti-perforin PE (S16009A, BioLegend, 154305, 1:100). Cells were acquired using a BD LSRFortessa flow cytometer (BD Biosciences) and analyzed on a FlowJo v.10 analysis platform (FlowJo, LLC).

RNA was extracted using the Aurum Total RNA kit (Bio-Rad), and cDNA was produced using iScript RT Supermix RT-qPCR (Bio-Rad) from 2 μg of total RNA in a 40 μl RT reaction. Quantitative real-time PCR was performed using the SSoFast EvaGreen Supermix (Bio-Rad) at an annealing temperature of 55°C. The final concentration of primers was 400 nM, and 30 ng (3 μl) of cDNA was used in a 10 μl qPCR reaction volume. Primer sequences for human ULBP1 (NM_025218), ULBP2 (NM_025217), MICA (NM_000247), MICB (NM_005931), and β-tubulin (housekeeping gene) mRNAs are listed in Supplementary Table 1 . The fold-change was calculated using the formula 2^(−ΔΔCt).

The projects TCGA-READ (rectal adenocarcinoma), TCGA-COAD (colon adenocarcinoma), and TCGA-BRCA (breast carcinoma) were downloaded using the R package TCGAbiolinks v2.25.3. Outlier samples were detected using arrayQualityMetrics before and after gene count normalization using edgeR. Samples were considered outliers and removed if they were marked as outliers before and after normalization using the same metrics or if they were marked as outliers by multiple metrics after normalization (COAD n=8; READ n=1; BRCA n=13). arrayQualityMetrics uses three metrics to mark outliers: 1) the total distance between a sample and all other samples, 2) the Kolmogorov-Smirnov test between a sample’s distribution and the pooled distribution of all other samples, and 3) Hoeffding’s statistic to determine sample independence. Highly variable genes were selected based on their median absolute deviations (MAD). For COAD and READ, genes with MAD ≥ the 55th most variable gene (MAD=0.033) were retained. For BRCA, genes whose MAD was ≥ the 50th most variable gene (MAD=0.022) were retained. Overall, between 27,300 and 30,321 genes remained for analyses. Differential expression analyses of the 11 candidate genes were performed using edgeR: ULBP1, ULBP2, ULBP3, MICA, MICB, CD274, RAET1L, RAET1G, HCST, CD276, and KLRK1. Differentially expressed genes had a Bonferroni Correction adjusted p < 0.05.

Statistical analyses were performed using the GraphPad Prism 9 software (GraphPad Software, San Diego, CA, USA). One-way ANOVA or two-way ANOVA tests with Dunnett’s or Holm-Sidak corrections were employed for multiple comparisons involving either a single variable or multiple variables, respectively. When comparing only two conditions for a single variable, a t-test was used with either Welch’s correction or Sidak’s correction. Data are presented as the mean ± SD unless otherwise indicated (*p<0.05, **p<0.005, ***p<0.0005, and ****p<0.0001).

Direct cytotoxicity of CuET was assessed in vitro using murine CRC cell lines MC-38 and CT-26 and human HCT116. MC-38 harbors a KRAS mutant and MSI/MMR-d and is a cell line model of an immune hot tumor. CT-26 do not express a mutation at the KRAS locus (wild-type) and are MSS/MMR-p, modeling an immune cold tumor phenotype. CuET inhibited cell viability in a dose-dependent with mean half-inhibitory concentrations (IC50) of 45.4nM and 68.2nM for MC-38 and CT-26, respectively ( Figures 1A, B ). In HCT116 cell lines, the parental cell line heterozygote for mutant KRAS at position G13D, displayed an IC50 of 54.8nM, the wild-type KRAS cell line had an IC50 of 54.0nM, and the homozygous mutated KRAS^G13D cell line had an IC50 of 48.3nM ( Figures 1C–E ). Additionally, CuET inhibited the survival and proliferation of all CRC cell lines in a dose-dependent manner, with significant inhibition occurring at 5nM, as demonstrated by the clonogenic assay ( Figures 1F, G ). Finally, the migration-invasion potential of the cells was assessed in a Boyden chamber, where CuET inhibited cell migration and invasion through a collagen matrix at doses ranging from 50 to 800nM in HCT116 KRAS^G13D cells and 100nM to 1µM in CT-26 cells ( Figures 1H, I ). Cell death by apoptosis was confirmed by PARP cleavage and XIAP degradation in HCT116 parental and HCT116 KRAS^G13D cells ( Figure 1J ).

In vivo safety assessment was performed by administering the albumin-CuET nanoparticle formulation (1 mg/kg) to 8-week-old BALB/c mice every three days over 17 days. Monitoring revealed no adverse events or deaths during treatment. CuET injections did not affect body weight ( Supplementary Figure 1A ), indicating that it was well tolerated in vivo. Blood biochemical parameters in BALB/c mice, including liver and renal functions, remained within the reference ranges and matched vehicle-treated and untreated animals, as evidenced by normal plasma AST, ALT, and creatinine levels ( Supplementary Table 2 ).

Histological examination revealed no internal organ toxicity, with no morphological changes in liver, kidneys, heart, or spleen compared to controls ( Supplementary Figure 2 ). Absence of vascular congestion, fatty changes, and inflammatory changes ruled out the possibility of hepatotoxicity. Renal toxicity was excluded because of the absence of glomerular atrophy and tubular necrosis. Myocardial hypertrophy, edema, or interstitial tissue fibrosis were not observed. Spleen assessment showed no hyperplasia or hypoplasia of the white or red pulp. Erythrocyte, platelet, and white blood cell counts remained consistent across the treatment and control groups ( Supplementary Figure 3 ). These findings collectively establish the excellent safety profile of CuET nanoparticles.

The therapeutic efficacy of albumin-CuET was evaluated in a syngeneic ectopic CT-26 CRC cold tumor model in BALB/c mice ( Figure 2A ). After three treatments, CuET significantly inhibited tumor growth by day 14, maintaining this difference until the end of the experiment on day 17 ( Figures 2B, C ). H&E staining revealed distinct morphological differences. Untreated tumors ( Figure 2E ) exhibited partial necrosis and poorly differentiated adenocarcinomas with pronounced vascularity and hemorrhage (blue arrows). Observational histological analysis identified numerous mitotic figures with elongated, aligned chromosomes (yellow arrows). In contrast, observational histological analysis of CuET-treated tumors showed an increase in vacuolated (red arrows), potentially senescent and dying, tumor cells ( Figure 2F ). TUNEL staining confirmed CuET-induced apoptosis, revealing increased apoptotic cells at the core of tumors in CuET-treated mice compared with vehicle-treated mice ( Figures 2G–L ).

Immunohistochemical analysis of excised tumors was performed to evaluate the cell populations within the tumor cores ( Figures 2M, N ). CuET treatment did not significantly enhance neutrophil or B-cell infiltration compared to vehicle treatment. However, CuET-treated mice exhibited significantly increased recruitment of macrophages and T cells to the tumor cores compared with the vehicle ( Supplementary Figure 4 ).

The therapeutic effect of CuET was assessed in MC-38 syngeneic ectopic and metastatic CRC hot tumor models in C57BL/6 mice ( Figure 3A ). Statistically significant inhibition of ectopic tumor growth was observed in the CuET group after five systemic treatments, with sustained differences in tumor volume until the end of the experiment on day 22 ( Figure 3B ) and reduced tumor weight at endpoint ( Figure 3C ). Interestingly, CuET-treated mice exhibited enlarged spleens ( Figure 3D ). Kaplan-Meier analysis indicated significantly prolonged survival in CuET-treated mice compared with vehicle-treated mice, with a median survival of 25 and 19 days, respectively ( Figure 3E ). In the metastatic model, CuET reduced liver metastatic lesions, occupying 15.73% of the liver area compared to 33.56% in controls after 14 days ( Figure 3F ). Overall, CuET inhibited tumor growth in syngeneic ectopic and metastatic models, and prolonged survival in mice.

Following the efficacy of CuET in inhibiting MC-38 metastatic tumor growth in the mouse liver, we explored the ex-vivo immune cytotoxicity of hepatocytes. CuET treatment of mice for six days increased ex-vivo cytotoxicity of hepatocytes against MC-38 cells over 24h ( Figure 3G ). This effect was accentuated in enriched cultures of NK cells ( Figure 3H ) and NKT cells ( Figure 3I ), demonstrating a lytic capacity approximately 2-fold higher than that of vehicle-treated mice. Significant differences were observed between 12h and 14h after co-culture (data not shown) and were sustained for 24h. Intracellular staining of mouse splenocytes stimulated ex-vivo with 1nM CuET revealed increased production of the effector cytokine perforin by CD8+ NK and T lymphocytes, compared to control ( Figures 3J, K ). Gating strategy can be found in Supplementary Figure 5 . Immunohistochemical staining of MC-38 tumors revealed positive staining for MICA (purple) and ULBP1 (green) ligands in both vehicle and CuET-treated groups ( Figures 3L, M ).

The ability of CuET to stimulate NK and T lymphocytes (effector) purified ( Supplementary Figures 6A, D ) from human PBMCs co-cultured with HCT116 and HT-29 cells (target) was assessed. Baseline cytotoxicity was seen in untreated NK and T cells ( Figures 4A–D ). Treatment with 1nM CuET significantly boosted NK and T cell cytotoxicity ( Figures 4A, B ). The same effect was observed when T cells were treated with CuET ( Figures 4C, D ). Additionally, target cell treatment with CuET maintained effector cell cytotoxicity at the same level as when only the effector cells were treated. Interestingly, an additive effect was observed when both effector and target cells were stimulated with CuET before co-culture, suggesting a two-fold effect: lymphocyte activation, enhanced immunogenicity, and targeting of cancer cells. These results prompted the characterization of the surface receptors on lymphocytes and their respective cancer cell ligands.

Flow cytometry analyses revealed increased NKG2D receptor expression on NK and T lymphocytes treated with 1nM CuET ( Figures 4E, F ) ( Supplementary Figures 6B, C, E, F ), as well as a significant increase in NKG2DLs MICA, ULBP1, and ULBP2 on human CRC cell lines HCT116 parental and HT-29, treated with 1nM CuET ( Figures 4G, H ). At the mRNA level, the modulation of NKG2D ligands in HCT116 parental cells under treatment with 1µM CuET induced a 3-fold and almost 4-fold increase in expression of ULBP1 and MICB, respectively, at 6h, and a 2.5-fold increase in ULBP2 expression at 24h ( Supplementary Figures 7A–D ). Time-dependent gene modulation was also observed in the HCT116 G13D cell line ( Supplementary Figures 7E–H ).

The importance of ligand expression in tumor cells previously stimulated with CuET prior to NK challenge was confirmed through antibody blocking of these ligands on cells, effectively decreasing the functional activity of NK cells in a dose-dependent manner, as evidenced by the inversely proportional decrease in cell cytotoxicity as antibody concentrations increased ( Figures 4I, J ). In HCT116 cells, the inhibition of NK cell cytotoxicity was higher when blocking MICA and ULBP1 (approximately 90% at 50ng) and less effective when blocking ULBP2 (75% at 50ng). In contrast, for HT-29 cells, inhibition of ULBP1 resulted in a 98% decrease in cytotoxicity at 50ng of antibody, whereas only a 50% and 43% decrease was observed when ULBP2 and MICA were blocked, respectively. The efficacy of antibody blocking with Fab fractions is shown in Supplementary Figure 8 . Correlation analysis between antibody blocking of NKG2D ligands and lytic units of NK cells against cancer cells was performed. Respectively, R² of 0.56 for MICA, 0.93 for ULBP1, and 0.88 for ULBP2 were obtained for HCT116 cells ( Supplementary Figure 8G ). Similarly, R² of 0.30 for MICA, 0.73 for ULBP1, and 0.60 for ULBP2 were obtained for HT29 ( Supplementary Figure 8H ).

Differential gene expression analysis across tumor samples and adjacent normal tissues from the COAD (colon adenocarcinoma), READ (rectum adenocarcinoma), and BRCA (breast carcinoma) projects of The Cancer Genome Atlas Program was performed after outlier removal and normalization ( Figure 5A ; Supplementary Figure 9 ). The analysis revealed an upregulation of NKG2D ligands in the tumor tissue ( Figures 5B–D ). Among 11 candidate genes, seven genes in the COAD dataset, four genes in the READ dataset, and five genes in the BRCA dataset were upregulated in the tumor tissues compared to normal tissue. The COAD tumor samples showed an increase between 4.5 and 11.3-fold for ULBP1/2/3/6 compared to normal tissues, and MICA and MICB had expression increases of 1.29 and 1.89-fold ( Supplementary Figure 10A ). READ tumors showed 3.03 to 6.06-fold increases in ULBP1/2/3 expression ( Supplementary Figure 10B ). BRCA tumors had 2.82 to 5.65-fold increases in ULBP1/2 expression, along with upregulation of RAET1L and MICB by 2.63 and 2.46-fold, respectively ( Supplementary Figure 10C ). Notably, MICA was significantly upregulated in tumors of the COAD but not BRCA dataset, while MICB was significantly upregulated in tumors of both COAD and BRCA. All three datasets saw a statistically significant upregulation of ULBP1/2 and CD276 in tumor samples versus adjacent normal tissue.